Ceph & OpenStack talk given @ OpenStack Meetup @ Bangalore, June 2015
0 likes•461 views
The document provides an overview of Ceph as a distributed storage system and its integration with OpenStack. It highlights Ceph's architecture, its capabilities such as self-healing and self-managing features, and its support for different storage types. Additionally, it discusses the advantages of using Ceph as unified storage for OpenStack, including optimizations for block, object, and ephemeral storage.
Ceph & OpenStack talk given @ OpenStack Meetup @ Bangalore, June 2015
- 1. 1 An Overview OpenStack BLR Meetup, June 2015 Deepak C Shetty (deepakcs) Ceph & OpenStack
- 2. 2 Why Ceph ?
- 3. 3 Ceph in OpenStack -2014 Source: http://superuser.openstack.org/articles/openstack-user-survey-insights-november-2014
- 4. 4 Ceph in OpenStack - 2015 Source: http://superuser.openstack.org/articles/openstack-users-share-how-their-deployments-stack-up
- 5. 5 What is Ceph ? ● Distributed storage system ● Algorithmic placement - CRUSH ● No single point of failure ● Self healing and self managing ● Runs on commodity hardware – No vendor lockin! ● Opensource – GPLv2 License – Community driven Probably, one of the best examples of SDS (aka Software Defined Storage)
- 7. 7 Ceph - Architecture - Not Production ready yet!
- 8. 8 RADOS COMPONENTS OSDs: 10s to 10000s in a cluster One per disk (or one per SSD, RAID group…) Serve stored objects to clients Intelligently peer for replication & recovery Monitors: Maintain cluster membership and state Provide consensus for distributed decision- making Small, odd number These do not serve stored objects to clients M
- 9. 9 OBJECT STORAGE DAEMONS FS DISK OSD DISK OSD FS DISK OSD FS DISK OSD FS btrfs xfs ext4 zfs? M M M
- 10. 10 CRUSH IS A QUICK CALCULATION RADOS CLUSTER OBJECT 1 0 0 1 0 1 1 0 1 0 0 1 1 1 0 1 1 0 0 1 0 1 1 0 1 0 0 1 1 1 0 1
- 11. 11 CRUSH: DYNAMIC DATA PLACEMENT CRUSH: Pseudo-random placement algorithm Fast calculation, no lookup Repeatable, deterministic Statistically uniform distribution Stable mapping Limited data migration on change Rule-based configuration Infrastructure topology aware Adjustable replication Weighting
- 12. 12 ACCESSING A RADOS CLUSTER APPLICATION M M M RADOS CLUSTER LIBRADOS OBJECT socket
- 13. 13 L LIBRADOS: RADOS ACCESS FOR APPS LIBRADOS: Direct access to RADOS for applications C, C++, Python, PHP, Java, Erlang Direct access to storage nodes No HTTP overhead
- 14. 14 THE RADOS GATEWAY M M M RADOS CLUSTER RADOSGW LIBRADOS socket RADOSGW LIBRADOS APPLICATION APPLICATION REST
- 15. 15 RADOSGW MAKES RADOS WEBBY RADOSGW: REST-based object storage proxy Uses RADOS to store objects API supports buckets, accounts Usage accounting for billing Compatible with S3 and Swift applications
- 16. 16 STORING VIRTUAL DISKS M M RADOS CLUSTER HYPERVISOR LIBRBD VM
- 17. 17 SEPARATE COMPUTE FROM STORAGE M M RADOS CLUSTER HYPERVISOR LIBRB D VM HYPERVISOR LIBRB D
- 18. 18 KERNEL MODULE FOR MAX FLEXIBLE! M M RADOS CLUSTER LINUX HOST KRBD
- 19. 19 RBD STORES VIRTUAL DISKS RADOS BLOCK DEVICE: Storage of disk images in RADOS Decouples VMs from host Images are striped across the cluster (pool) Snapshots Copy-on-write clones Support in: Mainline Linux Kernel (2.6.39+) Qemu/KVM, native Xen coming soon OpenStack, CloudStack, Nebula, Proxmox
- 20. 20 Ceph - Unified Storage for OpenStack
- 21. 21 Ceph as Unified Storage for OpenStack - Adv. ● No storage silos – Deploy/Manage 1 cluster with diff. pools ● Create image from volume and vice-versa optimizations ● Nova boot from volume optimizations ● Live migration ● Volume retype/migrate optimizations possible (WIP) ● Cinder Backup optimizations – Full and differential ● Cinder Volume replication (DR) made efficient via RBD mirroring (WIP)
- 22. 22 Ceph & OpenStack Storage- Summary ● Object Storage like Swift – Ceph RADOSGW as a drop-in replacement for OpenStack Swift ● Block Storage in Cinder – Ceph RBD pool for storing Cinder volumes ● Ephemeral Storage in Nova – Ceph RBD pool as backend for Ephemeral storage – Nova boot from volume ● Image Storage in Glance – Ceph RBD pool as a glance image store ● Backup target for Cinder-Backup – Ceph RBD pool as a backup target for Cinder – Backup / Restore cinder volumes to/from Ceph RBD pool ● File Storage in Manila (upcoming / future) – CephFS as a backend for Manila FS shares
- 23. 23 References ● Get the best configuration for your cloud – Devil is in the details – http://ceph.com/docs/master/rbd/rbd-openstack ● Ceph and openstack, current integration, roadmap (Vancouver summit prez) – http://www.slideshare.net/Red_Hat_Storage/open- stack-ceph-liberty
- 24. 24 Questions ? ● Disclaimer – Most (if not all) content for this prez taken from slideshare, youtube videos, ceph.com docs & other publicly available presentations.